Debug functions with Data Stream

You can send the log messages of your functions to an endpoint with a Data Stream stream, from Azion Console or with the Azion API. To trace which Rules Engine rules ran on a request instead, refer to Debug rules created with Rules Engine.

A function writes log messages with console.log. The Functions data source carries each message with its level, its source, and the ID of the request, and the Functions Event Collector template places them in one log line per message. For every variable of the data source, refer to Data sources and variables.

Prerequisites

Console

An Azion account with the Edit Data Stream permission. For the permissions, refer to Stream settings.
A function that writes log messages with console.log, running in an application that a workload serves. To create one, refer to Functions.
An endpoint that receives the log lines, such as an HTTP endpoint. For the endpoints a stream can send to, refer to Endpoints.
Access to Azion Console. To sign in, refer to How to access Azion Console.

Create the stream

The stream reads the Functions data source of one workload through a workload filter. Unlike sampling, a workload filter leaves your other streams active.

Console

To create the stream in Azion Console:

Open Data Stream

Access Azion Console > Data Stream.

Select + Stream

Name the stream

In the General section, enter a Name. For example: functions-debug.

Select the Functions data source

In the Input section, select Functions in Data Source.

Turn off Sampling

In the Transform section, turn off Sampling while Option is still All Current and Future Workloads, its starting value. A stream cannot carry sampling and a workload filter together.

Choose the workload

In the Transform section, set Option to Filter Workloads. In Available Workload, select the workload of the function and move it to Chosen Workload with the arrow. For the pick-list controls, refer to Associate workloads with a stream. If your Console names domains where this guide names workloads, the same control reads Filter Domains. The pick list then reads Available Domain and Chosen Domain.

Select the template

In the Render Template section, select Functions Event Collector in Template. The Data Set field shows the keys of each log line.

Select the endpoint

In the Output section, select your endpoint in Connector, and enter the fields it shows. Each endpoint has its own fields, listed on Endpoints.

Keep the stream active

In the Status section, keep Active turned on.

Select Save

The Console shows Your data stream has been created. The stream appears in the Data Stream list with functions_console in the Source column and the Active status.

Saving the stream checks the format of each field and does not contact the endpoint. A wrong URL or credential surfaces only when the stream sends. An activation takes effect after one to two minutes.

In the Deploy remote MCP servers use case, this section takes these values:

Setting	Value
Data Source	Functions. In the API, `functions_console`
Template	Functions Event Collector, whose `$log_message` variable carries each line the function writes. In the API, `86`
Option	Filter Workloads, with the workload that `azion deploy` created as the only chosen workload
Connector	Standard HTTP/HTTPS POST, with your log platform’s URL and the header it requires to accept the request

The stream becomes active within one to two minutes of saving it. Your log platform then receives one line per tool call, each carrying the $request_id of the request that made it.

In the Govern access to multiple AI models use case, this section takes these values:

Setting	Value
Data Source	Functions. In the API, `functions_console`
Template	Functions Event Collector, whose `$log_message` variable carries each line the function writes. In the API, `86`
Option	Filter Workloads, with the gateway’s workload as the only chosen workload
Connector	Standard HTTP/HTTPS POST, with your analytics platform’s URL and the header it requires to accept the request

The stream becomes active within one to two minutes of saving it. Your analytics platform then receives one model_call line per request, and one route_failed line per route that failed.

Confirm the delivery

Real-Time Events records every send of a stream, delivered or not, with the status code the endpoint returned. Send a few requests that run the function, then wait about a minute: a stream sends a batch every 60 seconds, or sooner when it reaches 2,000 log lines.

Console

To find the sends in Azion Console:

Open Real-Time Events

Access Azion Console > Real-Time Events.

Select the Data Stream data source

Read the latest sends

Each row is one send. Find the rows whose Endpoint Type matches your endpoint, such as HTTP_POST, and read their Status Code.

A Status Code of 200 means the endpoint accepted the batch. Streamed Lines gives the number of log lines in the batch.

A status other than 200 is the answer of the endpoint, and 503 means Data Stream found the endpoint unavailable. For the causes, refer to Troubleshoot Data Stream.

At the endpoint, each log line carries the keys of the Functions Event Collector template. Read these keys to debug the function:

logLevel holds the level of the message: ERROR, WARN, INFO, DEBUG, or TRACE. It carries the $log_level variable.
logMessage holds the message the function passed to the log function. It carries the $log_message variable.
messageSource reads CONSOLE for a message written with the console API, such as console.log, and RUNTIME for an error message.
requestID identifies the request, so the messages of one request share it.

To keep the variable names as keys, or to send fewer of them, select a custom template instead. For an example data set, refer to Templates and payload.

In the Deploy remote MCP servers use case, expect:

Every tool call is logged. After the get_order call, your log platform receives a line whose message holds "event":"tool_call" and "tool":"get_order".

In the Govern access to multiple AI models use case, expect:

Every call is logged. Your analytics platform receives a model_call line for each request sent to the gateway, with "team":"checkout-team". To read the lines of one request, open the Functions Console data source in Real-Time Events, where the ID variable groups the lines of a single request.

These checks confirm the Screen file uploads for malicious content use case.

Next steps

Data sources and variablesRead what each variable of the Functions data source holds, with an example value.FunctionsWrite and run the functions whose messages the stream collects.Create a custom templateChoose which Functions variables each log line carries, and under which keys.Troubleshoot Data StreamFind what to change when a send returns a status other than 200.Deploy remote MCP serversAn MCP server whose tool calls reach a log platform through a Functions stream.Govern access to multiple AI modelsA model gateway whose model calls and failed routes reach an analytics platform through a Functions stream.Screen file uploads for malicious contentAn upload screen whose verdicts reach the SIEM through a Functions stream.